Can Earlier BCG Vaccination Reduce Early Infant Mortality? A Randomised Trial

BCG, Infant Mortality

The purpose of this study is to determine whether BCG vaccination shortly after birth can reduce early infant mortality in a rural and an urban setting.

Background: BCG and oral polio vaccines (OPV) at birth are associated with beneficial non-specific effects, reducing neonatal mortality by more than what can be explained by prevention of the target diseases. BCG is recommended at birth, but is often given much later, especially in rural areas. In two RCTs in Guinea-Bissau, BCG-at-birth reduced neonatal mortality in low birthweight (\<2500g; LBW) children by 48% (95%CI: 18-67%) and in children with a birthweight \>2500g (NBW), OPV+BCG vs BCG was associated with a 32% (95%CI: 0-55%) lower mortality. WHO recommends home visits shortly after birth to reduce mortality, but vaccinations are not normally provided. If the vaccines indeed have profound effects on innate immunity and neonatal mortality in both LBW and NBW children many lives could be saved if BCG and OPV was provided earlier. Urban and rural clusters are randomised to home visits with and without vaccinations. All children participating in the study will be offered routine vaccines at village visits by the BHP team in the rural area. In the urban area, BCG and OPV will be provided at follow-up visits if the child has not yet received the vaccines. Thereby the study will provide earlier vaccination for all children. Hypothesis: BCG+OPV at birth provided at village visits shortly after birth will reduce early infant mortality by 40%. Methods: The study will be conducted in Biombo, Oio and Cacheu in rural Guinea-Bissau and in six suburban districts in the capital of Guinea-Bissau. In Guinea-Bissau home visits are not yet implemented as part of the routine program. Pregnant women will be offered to participate in the study at the time of pregnancy registration, which is conducted as part of the routine registration in the rural and urban health and demographic surveillance systems, respectively. Community key informants or mothers will communicate information on births to the BHP study team, and a study nurse will visit every new-born child shortly after a CKI or mother calls, if possible on the same day. Clusters will be randomised to receive immediate vaccination of their children shortly after birth or at the first visit by the BHP team in the rural area and at 2-months follow-up visits in the urban area. Statistical analyses: The primary analysis of early infant non-accidental mortality will be assessed on a PP analysis stratifying for factors used in the randomization (Region, pre-study mortality level (high/low)) and sex, thus allowing different baseline hazards for boys and girls. To account for clustering we will employ cluster-robust variance estimates. For the primary outcome, we will use Cox proportional hazards models, stratified for the above mentioned factors and with age as underlying time-scale. Deaths due to accidents will be censored. The effect of early vaccination will be assessed for the following secondary outcomes: * Non-accidental hospital admission * Severe morbidity (composite outcome of non-accidental mortality and non-accidental hospital admissions) * Consultations * Growth * Mid-upper-arm circumference * Weight-for-age z-score * BCG scarring * Cost-effectiveness of providing BCG and OPV at home visits Based on previous data from the rural HDSS in the areas where the current study will be conducted, the expected proportion of events (deaths and hospitalisation) between day 1 and the next home visit or 60 days of age, whichever comes first is 2.4% (unpublished data). The proportion of events are expected to be at least as high in the urban area. A recent trial in Ghana indicated that three home visits during the first week of life to promote essential new-born care practices and to weigh and assess children for danger signs was associated with an 8% (-12 to 25%) reduction in neonatal mortality. Based on pre-trial mortality data from the same rural clusters, the design effect is measured to be 1.43 (ratio of square of the standard errors for the cluster-adjusted/unadjusted HRs). Thus, in order to obtain 80% power to detect a reduction in early infant severe morbidity if the true reduction of BCG and OPV provided at home visits is larger than 40%, at least 6666 children need to be enrolled.

Guinea-Bissau